Author: Deric C.
Date: 2025-11-06T11:09:05-08:00
New Revision: 4c605e91244976fc7b746cfe8c0a90231b468ef7

URL: 
https://github.com/llvm/llvm-project/commit/4c605e91244976fc7b746cfe8c0a90231b468ef7
DIFF: 
https://github.com/llvm/llvm-project/commit/4c605e91244976fc7b746cfe8c0a90231b468ef7.diff

LOG: [HLSL] [DirectX] Invert the result of `firstbithigh` (#166419)

Fixes #145752 

This PR inverts the result of `firstbithigh` when targeting DirectX by
subtracting it from integer bitwidth - 1 to match the result from DXC.
The result is not inverted if `firstbithigh` returned -1 or when
targeting a backend other than DirectX.

Added: 
    

Modified: 
    clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
    clang/lib/Headers/hlsl/hlsl_intrinsic_helpers.h
    clang/lib/Headers/hlsl/hlsl_intrinsics.h
    clang/test/CodeGenHLSL/builtins/firstbithigh.hlsl
    clang/test/SemaHLSL/BuiltIns/firstbithigh-errors.hlsl

Removed: 
    


################################################################################
diff  --git a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h 
b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
index 208776eb7840e..2e2703de18cb1 100644
--- a/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_alias_intrinsics.h
@@ -1073,78 +1073,6 @@ float3 f16tof32(uint3);
 _HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_f16tof32)
 float4 f16tof32(uint4);
 
-//===----------------------------------------------------------------------===//
-// firstbithigh builtins
-//===----------------------------------------------------------------------===//
-
-/// \fn T firstbithigh(T Val)
-/// \brief Returns the location of the first set bit starting from the highest
-/// order bit and working downward, per component.
-/// \param Val the input value.
-
-#ifdef __HLSL_ENABLE_16_BIT
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint firstbithigh(int16_t);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint2 firstbithigh(int16_t2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint3 firstbithigh(int16_t3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint4 firstbithigh(int16_t4);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint firstbithigh(uint16_t);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint2 firstbithigh(uint16_t2);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint3 firstbithigh(uint16_t3);
-_HLSL_AVAILABILITY(shadermodel, 6.2)
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint4 firstbithigh(uint16_t4);
-#endif
-
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint firstbithigh(int);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint2 firstbithigh(int2);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint3 firstbithigh(int3);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint4 firstbithigh(int4);
-
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint firstbithigh(uint);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint2 firstbithigh(uint2);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint3 firstbithigh(uint3);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint4 firstbithigh(uint4);
-
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint firstbithigh(int64_t);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint2 firstbithigh(int64_t2);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint3 firstbithigh(int64_t3);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint4 firstbithigh(int64_t4);
-
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint firstbithigh(uint64_t);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint2 firstbithigh(uint64_t2);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint3 firstbithigh(uint64_t3);
-_HLSL_BUILTIN_ALIAS(__builtin_hlsl_elementwise_firstbithigh)
-uint4 firstbithigh(uint64_t4);
-
 
//===----------------------------------------------------------------------===//
 // firstbitlow builtins
 
//===----------------------------------------------------------------------===//

diff  --git a/clang/lib/Headers/hlsl/hlsl_intrinsic_helpers.h 
b/clang/lib/Headers/hlsl/hlsl_intrinsic_helpers.h
index c877234479ad1..3d8fe7ea701a6 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsic_helpers.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsic_helpers.h
@@ -148,6 +148,18 @@ template <typename T> constexpr T ldexp_impl(T X, T Exp) {
   return exp2(Exp) * X;
 }
 
+template <typename K, typename T, int BitWidth>
+constexpr K firstbithigh_impl(T X) {
+  K FBH = __builtin_hlsl_elementwise_firstbithigh(X);
+#if defined(__DIRECTX__)
+  // The firstbithigh DXIL ops count bits from the wrong side, so we need to
+  // invert it for DirectX.
+  K Inversion = (BitWidth - 1) - FBH;
+  FBH = select(FBH == -1, FBH, Inversion);
+#endif
+  return FBH;
+}
+
 } // namespace __detail
 } // namespace hlsl
 

diff  --git a/clang/lib/Headers/hlsl/hlsl_intrinsics.h 
b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
index 5ba5bfb9abde0..33ed14328ee8a 100644
--- a/clang/lib/Headers/hlsl/hlsl_intrinsics.h
+++ b/clang/lib/Headers/hlsl/hlsl_intrinsics.h
@@ -261,6 +261,67 @@ faceforward(__detail::HLSL_FIXED_VECTOR<float, L> N,
   return __detail::faceforward_impl(N, I, Ng);
 }
 
+//===----------------------------------------------------------------------===//
+// firstbithigh builtins
+//===----------------------------------------------------------------------===//
+
+/// \fn T firstbithigh(T Val)
+/// \brief Returns the location of the first set bit starting from the lowest
+/// order bit and working upward, per component.
+/// \param Val the input value.
+
+#ifdef __HLSL_ENABLE_16_BIT
+
+template <typename T>
+_HLSL_AVAILABILITY(shadermodel, 6.2)
+const inline __detail::enable_if_t<__detail::is_same<int16_t, T>::value ||
+                                       __detail::is_same<uint16_t, T>::value,
+                                   uint> firstbithigh(T X) {
+  return __detail::firstbithigh_impl<uint, T, 16>(X);
+}
+
+template <typename T, int N>
+_HLSL_AVAILABILITY(shadermodel, 6.2)
+const
+    inline __detail::enable_if_t<__detail::is_same<int16_t, T>::value ||
+                                     __detail::is_same<uint16_t, T>::value,
+                                 vector<uint, N>> firstbithigh(vector<T, N> X) 
{
+  return __detail::firstbithigh_impl<vector<uint, N>, vector<T, N>, 16>(X);
+}
+
+#endif
+
+template <typename T>
+const inline __detail::enable_if_t<
+    __detail::is_same<int, T>::value || __detail::is_same<uint, T>::value, 
uint>
+firstbithigh(T X) {
+  return __detail::firstbithigh_impl<uint, T, 32>(X);
+}
+
+template <typename T, int N>
+const inline __detail::enable_if_t<__detail::is_same<int, T>::value ||
+                                       __detail::is_same<uint, T>::value,
+                                   vector<uint, N>>
+firstbithigh(vector<T, N> X) {
+  return __detail::firstbithigh_impl<vector<uint, N>, vector<T, N>, 32>(X);
+}
+
+template <typename T>
+const inline __detail::enable_if_t<__detail::is_same<int64_t, T>::value ||
+                                       __detail::is_same<uint64_t, T>::value,
+                                   uint>
+firstbithigh(T X) {
+  return __detail::firstbithigh_impl<uint, T, 64>(X);
+}
+
+template <typename T, int N>
+const inline __detail::enable_if_t<__detail::is_same<int64_t, T>::value ||
+                                       __detail::is_same<uint64_t, T>::value,
+                                   vector<uint, N>>
+firstbithigh(vector<T, N> X) {
+  return __detail::firstbithigh_impl<vector<uint, N>, vector<T, N>, 64>(X);
+}
+
 
//===----------------------------------------------------------------------===//
 // fmod builtins
 
//===----------------------------------------------------------------------===//

diff  --git a/clang/test/CodeGenHLSL/builtins/firstbithigh.hlsl 
b/clang/test/CodeGenHLSL/builtins/firstbithigh.hlsl
index 368d652a6f779..51b0f81bea06a 100644
--- a/clang/test/CodeGenHLSL/builtins/firstbithigh.hlsl
+++ b/clang/test/CodeGenHLSL/builtins/firstbithigh.hlsl
@@ -1,161 +1,260 @@
 // RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
-// RUN:   dxil-pc-shadermodel6.3-library %s -fnative-half-type 
-fnative-int16-type \
-// RUN:   -emit-llvm -disable-llvm-passes -o - | FileCheck %s -DTARGET=dx
+// RUN:   dxil-pc-shadermodel6.3-library %s -fnative-half-type \
+// RUN:   -fnative-int16-type -emit-llvm -O1 -o - | FileCheck %s -DTARGET=dx \
+// RUN:   --check-prefixes=CHECK,DXCHECK
 // RUN: %clang_cc1 -finclude-default-header -x hlsl -triple \
-// RUN:   spirv-unknown-vulkan-compute %s -fnative-half-type 
-fnative-int16-type \
-// RUN: -emit-llvm -disable-llvm-passes \
-// RUN:   -o - | FileCheck %s -DTARGET=spv
+// RUN:   spirv-unknown-vulkan-compute %s -fnative-half-type \
+// RUN:   -fnative-int16-type -emit-llvm -O1 -o - | FileCheck %s -DTARGET=spv
 
 #ifdef __HLSL_ENABLE_16_BIT
 // CHECK-LABEL: test_firstbithigh_ushort
-// CHECK: call i32 @llvm.[[TARGET]].firstbituhigh.i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}i32 @llvm.[[TARGET]].firstbituhigh.i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub i32 15, [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq i32 [[FBH]], -1
+// DXCHECK-NEXT: select i1 %cmp.i, i32 -1, i32 [[SUB]]
+// CHECK-NEXT: ret i32
 uint test_firstbithigh_ushort(uint16_t p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ushort2
-// CHECK: call <2 x i32> @llvm.[[TARGET]].firstbituhigh.v2i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<2 x i32> 
@llvm.[[TARGET]].firstbituhigh.v2i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <2 x i32> splat (i32 15), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <2 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <2 x i1> %cmp.i, <2 x i32> splat (i32 -1), <2 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <2 x i32>
 uint2 test_firstbithigh_ushort2(uint16_t2 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ushort3
-// CHECK: call <3 x i32> @llvm.[[TARGET]].firstbituhigh.v3i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<3 x i32> 
@llvm.[[TARGET]].firstbituhigh.v3i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <3 x i32> splat (i32 15), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <3 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <3 x i1> %cmp.i, <3 x i32> splat (i32 -1), <3 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <3 x i32>
 uint3 test_firstbithigh_ushort3(uint16_t3 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ushort4
-// CHECK: call <4 x i32> @llvm.[[TARGET]].firstbituhigh.v4i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbituhigh.v4i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 15), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <4 x i32>
 uint4 test_firstbithigh_ushort4(uint16_t4 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_short
-// CHECK: call i32 @llvm.[[TARGET]].firstbitshigh.i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}i32 @llvm.[[TARGET]].firstbitshigh.i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub i32 15, [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq i32 [[FBH]], -1
+// DXCHECK-NEXT: select i1 %cmp.i, i32 -1, i32 [[SUB]]
+// CHECK-NEXT: ret i32
 uint test_firstbithigh_short(int16_t p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_short2
-// CHECK: call <2 x i32> @llvm.[[TARGET]].firstbitshigh.v2i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<2 x i32> 
@llvm.[[TARGET]].firstbitshigh.v2i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <2 x i32> splat (i32 15), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <2 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <2 x i1> %cmp.i, <2 x i32> splat (i32 -1), <2 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <2 x i32>
 uint2 test_firstbithigh_short2(int16_t2 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_short3
-// CHECK: call <3 x i32> @llvm.[[TARGET]].firstbitshigh.v3i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<3 x i32> 
@llvm.[[TARGET]].firstbitshigh.v3i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <3 x i32> splat (i32 15), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <3 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <3 x i1> %cmp.i, <3 x i32> splat (i32 -1), <3 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <3 x i32>
 uint3 test_firstbithigh_short3(int16_t3 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_short4
-// CHECK: call <4 x i32> @llvm.[[TARGET]].firstbitshigh.v4i16
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbitshigh.v4i16
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 15), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <4 x i32>
 uint4 test_firstbithigh_short4(int16_t4 p0) {
   return firstbithigh(p0);
 }
 #endif // __HLSL_ENABLE_16_BIT
 
 // CHECK-LABEL: test_firstbithigh_uint
-// CHECK: call i32 @llvm.[[TARGET]].firstbituhigh.i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}i32 @llvm.[[TARGET]].firstbituhigh.i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub i32 31, [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq i32 [[FBH]], -1
+// DXCHECK-NEXT: select i1 %cmp.i, i32 -1, i32 [[SUB]]
+// CHECK-NEXT: ret i32
 uint test_firstbithigh_uint(uint p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_uint2
-// CHECK: call <2 x i32> @llvm.[[TARGET]].firstbituhigh.v2i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<2 x i32> 
@llvm.[[TARGET]].firstbituhigh.v2i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <2 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <2 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <2 x i1> %cmp.i, <2 x i32> splat (i32 -1), <2 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <2 x i32>
 uint2 test_firstbithigh_uint2(uint2 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_uint3
-// CHECK: call <3 x i32> @llvm.[[TARGET]].firstbituhigh.v3i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<3 x i32> 
@llvm.[[TARGET]].firstbituhigh.v3i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <3 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <3 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <3 x i1> %cmp.i, <3 x i32> splat (i32 -1), <3 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <3 x i32>
 uint3 test_firstbithigh_uint3(uint3 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_uint4
-// CHECK: call <4 x i32> @llvm.[[TARGET]].firstbituhigh.v4i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbituhigh.v4i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <4 x i32>
 uint4 test_firstbithigh_uint4(uint4 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ulong
-// CHECK: call i32 @llvm.[[TARGET]].firstbituhigh.i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}i32 @llvm.[[TARGET]].firstbituhigh.i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub i32 63, [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq i32 [[FBH]], -1
+// DXCHECK-NEXT: select i1 %cmp.i, i32 -1, i32 [[SUB]]
+// CHECK-NEXT: ret i32
 uint test_firstbithigh_ulong(uint64_t p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ulong2
-// CHECK: call <2 x i32> @llvm.[[TARGET]].firstbituhigh.v2i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<2 x i32> 
@llvm.[[TARGET]].firstbituhigh.v2i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <2 x i32> splat (i32 63), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <2 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <2 x i1> %cmp.i, <2 x i32> splat (i32 -1), <2 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <2 x i32>
 uint2 test_firstbithigh_ulong2(uint64_t2 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ulong3
-// CHECK: call <3 x i32> @llvm.[[TARGET]].firstbituhigh.v3i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<3 x i32> 
@llvm.[[TARGET]].firstbituhigh.v3i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <3 x i32> splat (i32 63), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <3 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <3 x i1> %cmp.i, <3 x i32> splat (i32 -1), <3 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <3 x i32>
 uint3 test_firstbithigh_ulong3(uint64_t3 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_ulong4
-// CHECK: call <4 x i32> @llvm.[[TARGET]].firstbituhigh.v4i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbituhigh.v4i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 63), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <4 x i32>
 uint4 test_firstbithigh_ulong4(uint64_t4 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_int
-// CHECK: call i32 @llvm.[[TARGET]].firstbitshigh.i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}i32 @llvm.[[TARGET]].firstbitshigh.i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub i32 31, [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq i32 [[FBH]], -1
+// DXCHECK-NEXT: select i1 %cmp.i, i32 -1, i32 [[SUB]]
+// CHECK-NEXT: ret i32
 uint test_firstbithigh_int(int p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_int2
-// CHECK: call <2 x i32> @llvm.[[TARGET]].firstbitshigh.v2i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<2 x i32> 
@llvm.[[TARGET]].firstbitshigh.v2i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <2 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <2 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <2 x i1> %cmp.i, <2 x i32> splat (i32 -1), <2 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <2 x i32>
 uint2 test_firstbithigh_int2(int2 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_int3
-// CHECK: call <3 x i32> @llvm.[[TARGET]].firstbitshigh.v3i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<3 x i32> 
@llvm.[[TARGET]].firstbitshigh.v3i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <3 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <3 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <3 x i1> %cmp.i, <3 x i32> splat (i32 -1), <3 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <3 x i32>
 uint3 test_firstbithigh_int3(int3 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_int4
-// CHECK: call <4 x i32> @llvm.[[TARGET]].firstbitshigh.v4i32
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbitshigh.v4i32
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <4 x i32>
 uint4 test_firstbithigh_int4(int4 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_long
-// CHECK: call i32 @llvm.[[TARGET]].firstbitshigh.i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}i32 @llvm.[[TARGET]].firstbitshigh.i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub i32 63, [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq i32 [[FBH]], -1
+// DXCHECK-NEXT: select i1 %cmp.i, i32 -1, i32 [[SUB]]
+// CHECK-NEXT: ret i32
 uint test_firstbithigh_long(int64_t p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_long2
-// CHECK: call <2 x i32> @llvm.[[TARGET]].firstbitshigh.v2i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<2 x i32> 
@llvm.[[TARGET]].firstbitshigh.v2i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <2 x i32> splat (i32 63), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <2 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <2 x i1> %cmp.i, <2 x i32> splat (i32 -1), <2 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <2 x i32>
 uint2 test_firstbithigh_long2(int64_t2 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_long3
-// CHECK: call <3 x i32> @llvm.[[TARGET]].firstbitshigh.v3i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<3 x i32> 
@llvm.[[TARGET]].firstbitshigh.v3i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <3 x i32> splat (i32 63), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <3 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <3 x i1> %cmp.i, <3 x i32> splat (i32 -1), <3 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <3 x i32>
 uint3 test_firstbithigh_long3(int64_t3 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_long4
-// CHECK: call <4 x i32> @llvm.[[TARGET]].firstbitshigh.v4i64
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbitshigh.v4i64
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 63), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: ret <4 x i32>
 uint4 test_firstbithigh_long4(int64_t4 p0) {
   return firstbithigh(p0);
 }
 
 // CHECK-LABEL: test_firstbithigh_upcast
-// CHECK: [[FBH:%.*]] = call <4 x i32> @llvm.[[TARGET]].firstbituhigh.v4i32(<4 
x i32> %{{.*}})
-// CHECK: [[CONV:%.*]] = zext <4 x i32> [[FBH]] to <4 x i64>
-// CHECK: ret <4 x i64> [[CONV]]
+// CHECK: [[FBH:%.*]] = tail call {{.*}}<4 x i32> 
@llvm.[[TARGET]].firstbituhigh.v4i32(<4 x i32> %{{.*}})
+// DXCHECK-NEXT: [[SUB:%.*]] = sub <4 x i32> splat (i32 31), [[FBH]]
+// DXCHECK-NEXT: [[ICMP:%.*]] = icmp eq <4 x i32> [[FBH]], splat (i32 -1)
+// DXCHECK-NEXT: select <4 x i1> %cmp.i, <4 x i32> splat (i32 -1), <4 x i32> 
[[SUB]]
+// CHECK-NEXT: [[ZEXT:%.*]] = zext <4 x i32> {{.*}} to <4 x i64>
+// CHECK-NEXT: ret <4 x i64> [[ZEXT]]
 uint64_t4 test_firstbithigh_upcast(uint4 p0) {
   return firstbithigh(p0);
 }

diff  --git a/clang/test/SemaHLSL/BuiltIns/firstbithigh-errors.hlsl 
b/clang/test/SemaHLSL/BuiltIns/firstbithigh-errors.hlsl
index f99e606fc6562..1f70186c78ad9 100644
--- a/clang/test/SemaHLSL/BuiltIns/firstbithigh-errors.hlsl
+++ b/clang/test/SemaHLSL/BuiltIns/firstbithigh-errors.hlsl
@@ -12,7 +12,7 @@ int test_too_many_arg(int p0) {
 
 double test_int_builtin(double p0) {
   return firstbithigh(p0);
-  // expected-error@-1 {{call to 'firstbithigh' is ambiguous}}
+  // expected-error@-1 {{no matching function for call to 'firstbithigh'}}
 }
 
 double2 test_int_builtin_2(double2 p0) {


        
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

Reply via email to